Background/Aim: To evaluate the frequency and geographical variation in knee arthroscopy for adults (>25 years) with a concomitant diagnosis of osteoarthritis. Methods: This was a retrospective cohort study of hospital separations involving an elective knee arthroscopy in public and private hospitals in Victoria, Australia. Participants included patients receiving knee arthroscopies with a diagnosis code indicating osteoarthritis (OA) from 1 July 2008 to 30 June 2009. Records were excluded if the patient was under 25 years or their arthroscopy involved a ligament reconstruction. Crude rates per 100000 population and negative binomial regression offset by total knee arthroscopy volume were used to analyse differences by region. Results: There were 9620 arthroscopic procedures meeting the inclusion criteria. There were 5500 (57.2 ) admissions where the principal diagnosis was knee OA (gonarthrosis) and 3510 (36.5 ) where the principal diagnosis indicated a mechanical derangement and there was a primary or associated diagnosis of OA. When we examined the incidence rate ratios (IRR) by region, after adjustment for relevant factors and accounting for the total knee arthroscopy volume within each region, we identified significant variation in knee arthroscopy rates for patients with OA. The region with the highest adjusted IRR was Barwon South Western (IRR: 1.26, 95 confidence interval (CI): 1.16-1.36) and the region with lowest adjusted incidence rate ratio was the Gippsland region (IRR: 0.89, 95 CI: 0.80-0.98). Conclusions: We identified considerable geographical variation in arthroscopies for people with OA across Victoria. Further investigation is needed to understand whether this variation is a reflection of differences in OA prevalence, clinical decision-making or access.